R. A. Fisher, H. G. Thornton, and W. A. Mackenzie 351 



The counts therefore supply an example of the degree of accuracy 

 obtained with a gelatine medium, where a considerable source of variance 

 is produced by the occurrence of liquefying organisms on the plates. 



From the mass of data available, 100 sets of triplicate platings were 

 extracted. The expected and observed values of ■)^ in this series are 

 shown in Table XVIII. 



Table XVIII 



There is a marked deficiency below 1 , and an increasing excess above 3. 

 No distinct class of exceptionally high values can be detected, only three 

 values exceed 10, and none exceed 15. The causes of additional varia- 

 bility probably affect all observations in some degree, and are therefore 

 systematic rather than sporadic. The mean variance is about 50 per cent, 

 in excess of that due to random sampling. As in Cutler's 3-plate data 

 the departure from expectation is best shown by dividing the distribution 

 at the quintiles as in Table XIX. 



Table XIX 



X^ = 17-4 P = 0017 



